The macroeconomic price of AI

What happened

Equity markets remain near all-time highs, but beneath the surface, capital is quietly repositioning. Since late 2025, money has been flowing out of technology and into industrials, financials, consumer staples, energy. The equal-weight S&P 500 has been outperforming the headline index, meaning the average stock is doing better than the giants. Exxon, Coca-Cola, Walmart are at or near highs while  Software has suffered a -34% drawdown over twelve months. This is the sharpest non-recessionary decline in over thirty years for the sector, and yet every S&P 500 software company beat Q4 earnings. The market isn't reacting to bad results. It is repricing an entire category of business.

None of this constitutes a crisis. But it does constitute a pattern that is worth understanding before it becomes obvious to everyone.

Impact on our Investment Case

Why the software debacle is a macro question

The tech economy works as a loop. Startups buy cloud and SaaS tools. Cloud providers use that revenue to buy semiconductors. The revenues, valuations, and confidence generated at each layer feed the others. AI disrupts this loop from the inside: if a handful of AI platforms can do what thousands of SaaS companies were built for, an entire layer of customers disappears. And the ecosystem that depends on them contracts with it.

But the loop doesn't stop at tech. The knowledge workers most directly in AI's path (i.e. software engineers, legal professionals, financial analysts) are also consumers. In the US, personal consumption represents 68% of GDP. If companies start cutting headcount in anticipation of AI-driven efficiency, the impact spreads from tech to restaurants, retail, housing, services and every other business that depends on those paychecks.

There are early signs. US job openings have fallen to their lowest since 2017. Hiring has been frozen for six straight months. Nearly a million job cuts were announced in the first nine months of 2025. Crucially, Harvard Business Review found that many of these cuts are driven by the perception of what AI can do, not by actual deployment. The mechanism that could hurt the economy is fear, not displacement.

This is not without precedent. In 2000, technology represented roughly 33% of the S&P 500, the same share it commands today. The repricing that followed didn't stay in tech: it triggered a recession, pushed unemployment from 4% to 6%, and dragged consumer spending down with it. Tech's weight in the index collapsed from 33% to 14%. Today's tech leaders are profitable, cash-generative businesses, so this is unlikely to be a replay. But the pattern is clear: when a sector this large reprices, economic aftershocks follow.

Three channels to watch

If the adjustment deepens, these are the mechanisms through which it would spread.

Capex draining liquidity 

The largest technology companies are investing on a scale without modern precedent: roughly $400bn in 2025, and  $600–630bn projected for 2026, approximately 1.9% of US GDP. This spending validates that AI demand is real. But it is absorbing almost all of these companies' free cash flow, crowding out the share buybacks. For investors, this means one of the pillars under equity valuations (i.e., steady, large-scale buybacks from the biggest companies) is being quietly removed.

Private credit woes

Over the past few years, private lenders financed a wave of software company buyouts at peak valuations. JPMorgan estimates that software now represents roughly 16% of the loan books of publicly traded private credit funds (known as Business Development Companies, or BDCs), with potential losses of $22bn in a stress scenario. The stress is already visible in specific cases: Blue Owl Capital's tech-focused BDC saw 15.4% of its net assets withdrawn in January alone. If software valuations don't recover, these losses won't stay contained. Roughly $40bn of BDC assets overlap with public loans markets, creating a direct channel from private stress to public portfolios.

Fed chairman transition 

The nomination of Kevin Warsh as Fed Chair signals a shift. Warsh favors less balance sheet support from the central bank, with the Treasury taking a larger role in funding strategic spending directly. In simple terms: short-term rates may ease, but long-term borrowing costs lose the safety net of the Fed standing behind them. This matters because demand for capital is surging from all directions at once (namely AI infrastructure, government deficits, corporate refinancing) and all drawing from the same pool. For bond and equity investors alike, it means less liquidity support at exactly the moment when markets may need it most.

None of these channels is flashing red today. But they are interconnected, and they would reinforce each other if stress in one area triggered reassessment in the others.

The global dimension

This is not only an American story. China's DeepSeek demonstrated that algorithmic breakthroughs can match Western AI capabilities at a fraction of the cost, reshuffling assumptions about who wins the AI race and accelerating the pressure on incumbent business models globally. Europe, home to critical nodes in the AI supply chain (think ASML in semiconductor equipment, SAP in enterprise software), faces its own version of the adjustment: less exposed to the software repricing directly, but deeply connected to the capital expenditure cycle. When hyperscalers spend $600bn, that money flows across borders. And when the adjustment comes, so will the consequences.

What argues against a worst-case outcome

Three factors provide a counterweight.

Speed

After the dotcom bust, recovery took years because it depended on building physical infrastructure. That work couldn't be rushed. AI doesn't face the same constraint. DeepSeek and Google's 4.5x efficiency gains show that progress can leapfrog hardware. A difficult few quarters, rather than a multi-year grind, is the more plausible path.

Fiscal support

In the US, America's 250th anniversary in July and midterm elections in November create strong political incentives for economic intervention. In Europe, the push toward strategic autonomy in technology and defense is unlocking spending that has been constrained for years. In China, the government has repeatedly shown its willingness to act when growth falters. The global policy apparatus is primed to cushion the adjustment, even if it cannot prevent it entirely.

Real demand

This is the critical difference from 2000. The dotcom bust revealed that much of the "demand" was circular and illusory. Today, more than a billion people use AI tools every week. Cloud revenues are growing at double-digit rates across every major platform. Hyperscaler capital spending is still accelerating with no sign of plateauing. The demand is broad, measurable, and grounded in real productivity gains, not projections.

Our Takeaway

The purpose of this analysis is to map the channels through which AI disruption could move from a sector story to a macro one, so that the pattern is recognizable when it emerges. The indicators to watch are the labor market, private credit fund flows, and the incoming Fed chair's early policy signals.

The deeper point is about what comes next. Every major infrastructure buildout creates an asymmetry: the companies that bear the cost of building are not always the ones that capture the value. After the dotcom bust, the biggest winners were not the telecoms that laid the fiber or the hardware companies that survived the crash. They were Netflix, Airbnb, and the entire app economy. In other words, businesses that ran on top of cheap infrastructure built at someone else's expense. The same dynamic is taking shape in AI. Hundreds of billions are being poured into computing power, data centers, and foundational models. The companies that will benefit most from this investment are the ones that will use that computing power to transform how businesses operate, how professionals work, and how consumers interact with services, without bearing all the capital cost. That is the application layer. And the opportunity to invest in it is being created right now, as the market reprices the sector indiscriminately.

The transition will not be seamless. But the economy that emerges on the other side will operate at a higher level.